Hong Kong · Software engineering and technical consulting

We build the applications, systems and agents our clients run in production.

Sweet Intelligence Limited is a technology research and development company with its office in Central, Hong Kong. We write mobile, web and backend software, we build LLM agents on client data, and we advise engineering teams on architecture, code quality and technical risk.

App development System development Agent development Research and technical consulting

What we build

Four lines of work, each delivered by the engineers who scoped it. Every engagement produces running software in a repository the client owns, together with the documentation needed to operate it.

01

App development

We build iOS and Android applications and the web applications that sit alongside them. The work covers interface implementation, offline and sync behaviour, release builds, and submission to the App Store and Google Play.

Business logic is covered by automated tests before a build reaches a store.

Artefacts: repository, CI pipeline, test suite, signed release builds, store submission notes.

02

System development

We build backend services, APIs and data pipelines, and we integrate them with the systems a client already runs. Internal platforms, scheduled jobs, queues and reporting databases are part of the same scope.

Schema migrations are written to run forwards and to be reversed if a release is rolled back.

Artefacts: repository, API documentation, migration scripts, monitoring dashboards, deployment runbook.

03

Agent development

We build LLM agents and copilots that retrieve from client data, call internal tools, and return the work to a named person at defined review points. Each agent ships with an evaluation harness and a fixed test set, so a prompt or model change can be measured before it goes live.

Guardrails, rate limits and logging of prompts, tool calls and outputs are part of the build.

Artefacts: repository, evaluation harness and test set, prompt and tool specifications, observability dashboards.

04

Research and technical consulting

We review architecture and code, and we build prototypes to settle technical questions that a document cannot answer. We also carry out technical due diligence on software a client is acquiring or inheriting.

For longer engagements we place engineers inside a client team for a fixed period to work on that team's own backlog.

Artefacts: architecture note, prototype repository, findings memorandum, prioritised remediation list.

How a project runs

Projects run in short iterations, and each iteration finishes with a build a client can open and use. The durations below are the estimates we work to, and each one is confirmed in the scoped specification.

STAGE 01

Technical discovery and estimate

We read the existing code and data, and we talk to the people who maintain them, so that constraints, integrations and unknowns are written down rather than assumed.

The output is a scoped specification with an estimate for each component.

Typically 2 weeks

STAGE 02

Architecture and a working skeleton

We agree the stack, the data model and the interfaces between services, then stand up a deployable skeleton with continuous integration, environments and logging from the first commit.

A staging environment is live and reachable by the client at the end of this stage.

2 to 3 weeks

STAGE 03

Build in two week iterations

Features are built in two week iterations, each ending with a deployment to staging and a written changelog of what changed and what is still open.

Code review and the automated test suite run on every pull request before it is merged.

6 to 10 weeks to a first release

STAGE 04

Release, monitoring and hand-over

We run the production release, watch the first weeks of real traffic, and fix what the logs and error reports show.

Hand-over is complete once an engineer on the client side has deployed the system themselves using the runbook we wrote.

2 weeks, with support by agreement

Engineering practice

The team works across mobile, web, backend and model tooling, and chooses the stack that fits the systems a client already operates. The list below covers the technologies and practices we work with regularly.

TypeScript React React Native Swift Kotlin Node.js Python PostgreSQL Redis REST and GraphQL APIs AWS Cloudflare Docker GitHub Actions OpenAI and Anthropic APIs Vector search Retrieval pipelines Evaluation harnesses Structured logging and tracing Observability dashboards Code review Automated tests Infrastructure as code

Where a client already runs a different stack, we work inside it and say so in the estimate.

Sweet Intelligence Limited

Sweet Intelligence Limited is a technology research and development company registered in Hong Kong. The team builds software and advises client engineering teams from an office in Central, and written enquiries reach us by post at the address listed here.

Entity
Sweet Intelligence Limited
Type
Hong Kong limited company
Incorporated
21 May 2025, Hong Kong
Company number
78185856
Office
Flat/Rm 2108B, Champion Tower
Three Garden Road, Central, Hong Kong